Remove of scale parameter of eigen_centrality()

[szhorvat]

In C/igraph 1.0, the scale parameter of eigen_centrality() will be removed and the function will always behave as if scale=TRUE were used. Details are at igraph/igraph#2446

This issue is to start preparing for this change on the R side, probably by deprecating the scale parameter.

Note that no real functionality is lost: scale=FALSE doesn't guarantee anything about any kind of scaling/normalization, while scale=TRUE uses a specific normalization. People should never have relied on scale=FALSE, so warning those who do soon is a good thing.

CC @maelle

[maelle]

@szhorvat do you agree that we should

• message if the argument is present and TRUE
• already error if it is present and FALSE since it seems really problematic to think scale=FALSE does something?

[szhorvat]

@maelle I'm sorry, somehow I missed this message when I responded in the PR.

I think at this point we should NOT error for scale=F. There is nothing wrong with eigen_centrality(scale=F), it just doesn't guarantee any specific kind of normalization. It was an error in the documentation to claim that it did.

---

But there is also centr_eigen() and centr_eigen_tmax(). It turns out that scale=F made no sense with these. Now there is a warning issued by the C core when FALSE is passed, and it's treated as TRUE anyway. This change wasn't released so far. See igraph/igraph#2674. This applied only to centr_eigen() and centr_eigen_tmax(), but not to eigen_centrality().